Codeword stabilized quantum codes Andrew Cross, Graeme Smith, John A. Smolin, and Bei Zeng CWS codes for short 11111 quant-ph/0708.1021 Venn Diagram All codes CWS stabilizer classical Some notation [[n,k,d]] additive quantum code: n raw qubits, k protected qubits, distance d ((n,K,d)) quantum code: n raw qubits, K protected states, distance d For an additive code K=2k ((9,12,3)) Our work was inspired by the ((9,12,3)) code of Yu, Chen, Lai and Oh quant-ph/0704.2122. This was the first nonadditive quantum code with distance > 2 that outperforms any known additive code (and even any possible additive code). Stabilizers We consider only the Pauli group Pauli group: I,X,Z,Y and tensor products written like XIZZIYZ Stabilizer for an [[n,k,d]] code is a set of commuting members of this group generated by S1...Sn-k There are also logical operators X1...Xk and Z1...Zk These also commute with the stabilizers A stabilizer state is an [[n,0,d]] code, i.e. it is the +1 eigenstate of a maximal abelian subgroup of the Paulis having n generators and no logical operators. CWS codes are characterized by two objects: Standard form CWS codes are defined by two objects The stabilizer tells you which errors the classical code must correct Error detection conditions (general) Error detection conditions Codewords should not be confused Error should be detected Last two codewords are “immune” to the error---degeneracy condition Graph States Graph states are stabilizer states which have stabilizer generators each with a single X and Z's on the nodes to which they're connected. IXZZI 3 IZXZZ 2 XIIII 1 4 IIZZX 5 IZZXZ Standard Form Theorem: Any codeword stabilized code is locally equivalent to one with a graph state stabilizer and word operators consisting only of Z's and including the identity. We call this standard form. Proof ingredients: 1. any stabilizer state is locally clifford equivalent to a graph state. (proved elsewhere) basically row-reduction 2. This results in new codeword operators, still products of Paulis. 3. Any X's in the new codewords can be eliminated by multiplying by stabilizer elements from the graph state. Since these each have a single X, this is straightforward. X-Z rule (lemma) On a graph state X errors are equivalent to (possibly multiple) Z errors. We call these the induced errors. Si has only one X on bit i so the X's cancel X-Z rule Z X Z Z Z Z Z Y Y=XZ Z Errors detection conditions (standard form) Since all induced errors are Z's, things are essentially classical (degeneracy) Relation to stabilizer codes Furthermore, whenever the word operators form a group, a CWS code IS a stabilizer code Antideluvian [[5,1,3]] code BDSW96 'the big paper' code also in LMPZ96 [[5,1,3]] codewords all parity 0 string with some collection of signs, and the same with 0 and 1 interchanged [[5,1,3]] Stabilizer XZZXI IXZZX Generators of the stabilizer XIXZZ ZXIXZ XXXXX logical X ZZZZZ logical Z Can be made into a CWS code by adding in XXXX to the stabilizer, and using 00000 and ZZZZZ as the codeword operators Illuminating? On a ring To correct single errors, need to detect double errors Z If the codewords are 00000 and 11111 a Z nondetectable error would have to be weight 5 The X-Z rule tells us all single errors on Z Z the ring are weight 1, 2, or 3 Z these cancel We need one weight 3 and one weight 2 But the only weight 2 errors are nonadjacent ((5,6,2)) Rains, Harden, Shor, Sloane code “The symmetries discussed above generate a group of order 640. There is an additional symmetry which can be described as follows: First, permute the columns as k->k^3, that is exchange qubits 2 and 3. Next, for each qubit negate one of the Pauli matrices and exchange the other two, where the Pauli matrices negated are Z, Y, X, X, Y, respectively. This increases the size of the symmetry group to 3840. This group acts as the permutation group S5 on the qubits. This is the full group of symmetries of the code. That is, the full subgroup of the semidirect product of S5 and PSU25 that preserves the code [10].” ((5,6,2)) code 00000 11010 01101 10110 01011 10101 Since weight 3 induced errors are adjacent, the weight can't change by 3 so none of these can be transformed into 00000. Since weight 2 errors are non-adjacent, they can't be transformed amongst each other. ((9,12,3)) 9-ring ((10,18,3)) If rings are so great, why not a bigger one? 10-ring Linear programming bound is ((10,24,3)) ((10,20,3)) If one ring is good, two must be better Linear programming bound is ((10,24,3)) Big search problem For all graphs of size n, search for best classical code of a given distance. Super-exponential Turns out to be quite doable for n up to 10 or maybe 11 ((10,24,3)) code meeting linar programming bound: Yu, Chen and Oh, 0709.1780 ((10,24,3)) ((10,24,3)) code is unique and meets linear programming bound Almost the simplest nontrivial nonembeddable in 3D graph where edges represent orthogonality conditions Encoding circuits (Bei’s slide) U(G,C ) = GC ci Classicl Encoder C: C i = ci = X 00...0 ⊗n Gragh Encoder G: G 00..0 = G G = ∏P( j,k)H ( j,k ∈E ) ⊗n ci GC i = G ci = ∏ P( j,k ) H X 00...0 ( j,k∈E) ci ⊗n ci ⊗n = ∏ P( j,k )Z H 00...0 = Z ∏ P( j,k ) H 00...0 ( j,k∈E) ( j,k∈E) = Z ci G Future Work Find more codes! Particularly of higher distance Generalize to higher dimensions Looi, Yu, Gheorghiu and Griffiths 0712.1979 Understand strange classical error models CodewordCodeword StabilizedStabilized QQuantumuantum CodesCodes Isaac Chuang1, Andrew Cross1,2, Graeme Smith2, John Smolin2, and Bei Zeng1 1Massachusetts Institute of Technology, Cambridge, MA 02139, USA; 2IBM T. J. Watson Research Center, Yorktown Heights, NY 10598, USA Abstract -- We present a unifying approach to quantum error General Construction The CWS-MAXCLIQUE Algorithm correcting code design that encompasses additive (stabilizer) codes, as Given G and C, our CWS code This criterion leads to a necessary and Once the graph has been chosen, the problem of finding a quantum well as all known examples of nonaddtive codes with good parameters. is spanned by basis vectors sufficient condition for detecting errors code is reduced to a search for classical codes. There are no handy We consider an algorithm which maps the search for quantum codes in w = w G in that for all E " systematical constructions in classical coding theory to deal with codes our framework to a problem of identifying maximum cliques in a graph. l l " " "i # j w + Ew $ ±S, "i w + Ew # ±S or of such perhaps exotic error patterns induced by our approach. But we We use this framework to generate new codes with superior parameters where the graph state |G> is i j i i do know the algorithm of finding the best K for a given n,d has a stabilized by the stabilizer S + to any previous known. In particular, we find ((10,18,3)) and ((10,20,3)) and "i wi Ewi $ S or natural encoding into the problem of MAXCLIQUE. r l + codes. ! Sl = X l Z ! "i wi Ewi $ %S We present a MAXCLIQUE algorithm for finding our CWS quantum where r is the lth row of the code for the largest possible dimension K for a given n,d, and G. This Motivation & Idea l ! Due to this condition, all errors become graph’s adjacency ma!trix , and Zs, we c!an think the error model as CWS-MAXCLIQUE algorithm (Algorithm 3) proceeds in three simple Quantum error correction codes play a central role in quantum w = Z c l steps: ! l classical, albei!t c onsisting of strange computation and quantum information. While considerable understanding The first step finds the elements of The CWS-MAXCLIQUE are Pauli operators defined by multibit errors. We define this # 0 1 0 0 1& has now been obtained for a broad class of quantum codes, almost all of ! % ( algorithm for 1 0 1 0 0 translation to classical errors by the Cl G ( " ) and D G ( " ) . D G ( " ) is a set of this has focused on stabilizer codes, the quantum analogues of classical the classical codeword constructing the ((5,6,2)) % ( classical bit strings defined by " = % 0 1 0 1 0( additive codes. However, such codes are strictly suboptimal in some function code in [3]. Starting from % ( cl " C 0 0 1 0 1 ! % ( settings---there exist nonadditive codes which encode a larger logical space The quantum error correction n c the five qubit ring graph 1 0 0 1 0 n {c " {0,1} [Z ,E] # 0 for some E " $ % S} $ ' than possible with a stabilizer code of the same length and capable of criterion Cl E Z v X u v u r ! ! ! with adjacency matrix G ( = ± ) = "( )l l tolerating the same number of errors. There are only a handful of such + l=1 which is nonempty only if the CWS The code is nondegenerate. The set G wi Ew j G = cE"ij examples, and their constructions have proceeded in an ad hoc fashion, code (G,C) is degenerate.